115 lines
4.3 KiB
Python
115 lines
4.3 KiB
Python
import tensorflow as tf
|
|
import numpy as np
|
|
import random
|
|
import pickle
|
|
from PIL import Image
|
|
from qrcode import make as makeqr
|
|
from dnnlib import tflib
|
|
import time, os, hashlib
|
|
|
|
|
|
def main():
|
|
# Define global variables.
|
|
seed = random.randint(0,10000000)
|
|
available_charaters = {'Anmicius', 'Camil', 'Grey', 'King', 'Ray', 'Godrose'}
|
|
|
|
|
|
# Select charater and input seed.
|
|
selected_character = 'Ray'
|
|
while selected_character not in available_charaters:
|
|
selected_character = input('Type in the character you want to draw, e.g. \"Anmicius\" and \"Ray\" (no quotes).\n')
|
|
if selected_character not in available_charaters:
|
|
print('You typed in a character that is not available or you made a misspell, try agian.')
|
|
|
|
seed_str = ''
|
|
if seed_str != '':
|
|
if seed_str.isdigit():
|
|
seed = int(seed_str.encode('utf-8'))
|
|
else:
|
|
seed = int(hashlib.sha256(seed_str.encode('utf-8')).hexdigest(), 16) % 10**8
|
|
|
|
print('INFO: Setting up variables...')
|
|
tflib.init_tf()
|
|
rnd = np.random.RandomState(seed)
|
|
fmt = dict(func=tflib.convert_images_to_uint8, nchw_to_nhwc=True)
|
|
|
|
print('INFO: Loading pretrained model...')
|
|
Gs = pickle.load(open('models/network-%s-gs.pkl' % selected_character, 'rb'))
|
|
latents = rnd.randn(1, Gs.input_shape[1])
|
|
|
|
print('INFO: Generating...')
|
|
images = Gs.run(latents, None, truncation_psi=0.7, randomize_noise=True, output_transform=fmt)
|
|
|
|
im = Image.fromarray(images[0], 'RGB')
|
|
qr = makeqr('This is an image automatically generated by Aotu Draw Bot by Rand0mZ.LiCloud provides computing resources. Seed: %d' % seed)
|
|
w, h = im.size
|
|
qw, qh = qr.size
|
|
if qw > w:
|
|
qr = qr.resize((w, w))
|
|
elif qh > h:
|
|
qr = qr.resize((h, h))
|
|
qw, qh = qr.size
|
|
|
|
imd = im.load()
|
|
for i in range(w):
|
|
for j in range(h):
|
|
d = imd[i, j]
|
|
imd[i, j] = d[:-1] +((d[-1] | 1) if qr.getpixel((i%qw, j%qh)) else (d[-1] & ~1),)
|
|
|
|
print('Done!')
|
|
|
|
save_name = '%s_%d.png' % (selected_character, seed)
|
|
print('INFO: Saving %s' % save_name)
|
|
output_dir = os.path.join(os.path.dirname(os.getcwd()) , 'ARAGS/Ray')
|
|
if not os.path.isdir(output_dir):
|
|
os.mkdir(output_dir)
|
|
im.save(os.path.join(output_dir, save_name))
|
|
print('INFO: Image %s is saved in directory.' % save_name)
|
|
print('INFO: All processes has done!')
|
|
print('Thank you for using this software and obeying the terms of use above.')
|
|
time.sleep(3)
|
|
|
|
|
|
def generate_image(model, save_path, selected_character, seed, amount):
|
|
tflib.init_tf()
|
|
|
|
print('INFO: Loading pretrained model...')
|
|
Gs = pickle.load(open(model, 'rb'))
|
|
|
|
if not os.path.isdir(save_path):
|
|
os.mkdir(save_path)
|
|
|
|
for i in range(1, amount + 1):
|
|
print('INFO: Generating image %d' %i)
|
|
rnd = np.random.RandomState(seed)
|
|
fmt = dict(func=tflib.convert_images_to_uint8, nchw_to_nhwc=True)
|
|
latents = rnd.randn(1, Gs.input_shape[1])
|
|
images = Gs.run(latents, None, truncation_psi=0.7, randomize_noise=True, output_transform=fmt)
|
|
|
|
im = Image.fromarray(images[0], 'RGB')
|
|
qr = makeqr('This is an image automatically generated by Aotu Draw Bot CLI by Rand0mZ hence this image is not for commercial propose. Seed: %d' % seed)
|
|
w, h = im.size
|
|
qw, qh = qr.size
|
|
if qw > w:
|
|
qr = qr.resize((w, w))
|
|
elif qh > h:
|
|
qr = qr.resize((h, h))
|
|
qw, qh = qr.size
|
|
|
|
imd = im.load()
|
|
for i in range(w):
|
|
for j in range(h):
|
|
d = imd[i, j]
|
|
imd[i, j] = d[:-1] +((d[-1] | 1) if qr.getpixel((i%qw, j%qh)) else (d[-1] & ~1),)
|
|
|
|
print('Done!')
|
|
|
|
save_name = '%s_%d.png' % (selected_character, seed)
|
|
print('INFO: Saving %s' % save_name)
|
|
im.save(os.path.join(save_path, save_name))
|
|
print('INFO: Image %s is saved to %s.\n' % (save_name, save_path))
|
|
seed += i - 1
|
|
|
|
if __name__ == "__main__":
|
|
main()
|